Dissertation > Excellent graduate degree dissertation topics show

Research on Structure Learning of Bayesian Classifier Based on Genetic Algorithms

Author: JiangWangDong
Tutor: LinShiMin
School: Guangxi Normal University
Course: Computer Software and Theory
Keywords: Bayesian networks Structure Learning Bayesian classifier Genetic Algorithms MATLAB application
CLC: TP18
Type: Master's thesis
Year: 2005
Downloads: 735
Quote: 6
Read: Download Dissertation

Abstract

In the past two decades, the world economy has led to the rapid development of information technology, the rapid popularity of Internet technology, the capacity to collect data to greatly improved, making it possible to obtain increasingly large scale of data and need to be addressed. The face of the \Classification is a very important task in data mining, and its aim is to identify the classification function or classification model. The classification of some machine learning methods, such as the decision tree method, rule induction methods, neural network, genetic algorithm, ant algorithm. In many classification methods, Bayesian networks as an effective way of knowledge representation and probabilistic reasoning model, is a powerful graphical decision analysis tools to deal with uncertain information. Firm theoretical basis, the nature of knowledge representation, a flexible reasoning ability and decision-making mechanism, by more and more attention. In recent years, Bayesian network-based data mining to obtain good results, become a hot topic. The Bayesian method is based on Bayes' theorem developed for the elaboration and statistical problems. In order to establish the Bayesian network for classification called Bayesian classifier. Bayesian classifier is a special form of Bayesian networks, selection and status of the variables have been determined, the attribute node known class node unknown. The Bayesian classifier family has three types of common classification: Naive Bayes classifier NBC, tree expansion of naive Bayes classifier of TANC and a Bayesian network classifier BNC. Bayesian classifier learning including structure learning, parameter learning and maximum a posteriori probability class nodes reasoning. Fully Bayesian network structure learning is an NP-hard problem, researchers generally use the approximate method to learn, Duda put forward the Naive Bayes NB, Friedman Tree Augmented Na?ve Bayes TAN structure; Keogh the SP structure; Huajie Zhang proposed SN architecture; FAN Peter Lucas proposed structure; Cheng put forward BAN and GBN two kinds of network structure; SHI Hongbo TAN structure has been optimized. Bayesian classifier structure learning they have achieved good results. More optimized structure in a short period of time has been the key issues for everyone. Genetic Algorithm (Genetic Algorithm) is to simulate the natural process of biological evolution and mechanisms for solving extremal problems of a self-organizing, adaptive artificial intelligence technology. It comes from the natural evolutionary theory of Darwin and Mendel's theory of genetic variation, has a solid biological basis. The genetic algorithm is a global search optimization algorithm, which is obtained by simulating the process of biological evolution, the global optimal solution. The introduction of genetic algorithm to improve the Bayesian network structure learning problem to be solved in this article. The main work of this paper are as follows: (1) summarized the theoretical framework of the Bayesian network, a brief discussion of Bayesian network structure learning algorithm. (2) on the basis of Ze-Kai Cheng et al. [78, 79] using MATLAB language based on BNT (Bayesian Networks Toolkit) to construct Bayesian classifier experimental platform MBNC (Bayesian Networks Classifier using MATLAB), extended the MBNC experimental platform genetic algorithm module, the pre-processing of the data, based on genetic algorithm Bayesian classifier structure learning algorithm, so as to realize a variety of classification of the family of Bayesian classifier based on genetic algorithm. (3) For the genetic algorithm is introduced Bayesian structure learning, in-depth study of the genetic algorithm and MATLAB-based implementation, focusing on the integer coded genetic algorithm-based and TSP problem solving; efficient design for TSP left

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Automated basic theory > Artificial intelligence theory
© 2012 www.DissertationTopic.Net Mobile